Exact Holographic Tensor Networks -- Constructing CFT$_D$ from TQFT$_{D+1}$

Lin Chen,Hao-Chen Zhang,Kai-Xin Ji,Shen Chen,Ruo-shui Wang,Xiang-dong Zeng,Ling-Yan Hung
DOI: https://doi.org/10.48550/arxiv.2210.12127
2022-01-01
Abstract:In this paper, inspired by [1-3], we proposed a class of lattice renormalization group (RG) operators, each operator determined by a topological order $T$ in $D+1$ space-time dimensions. Taking the overlap between an eigenstate $\langle\Omega|$ of the RG operator with the ground state wave-function $|\Psi\rangle$ of $T$ (i.e. $\langle\Omega|\Psi\rangle$) gives rise to partition functions of conformal (including topological) theories in $D$ dimensions with categorical symmetry related to $T$, realizing a holographic relation discussed in the literature explicitly. We illustrate this in explicit examples at $D=1,2,3$. Exact eigenstates of the RG operator can be solved explicitly from (higher) separable Frobenius algebra of the (higher) input fusion category $\mathcal{C}$ defining the lattice model of $T$, and they give the $D$ dimensional symmetric TQFTs. Eigenstates corresponding to actual conformal theories describe phase transitions between these topological fixed points. The critical points can be searched numerically and we demonstrate that known critical couplings of $SU(2)_k$ integrable lattice models are numerically recovered from our procedure, alongside a curious tricritical point that we found at $k=4$. We demonstrate that the 2+1 D Ising model can also be obtained as a strange correlator with the associated 4D topological order being the 4D toric code. The numerical procedure that we devise to search for the 3D critical temperature is a novel tensor renormalization group algorithm, that fully harnesses the algebraic and geometric properties of the RG operator. Finally since the RG operator is in fact an exact analytic holographic tensor network, we compute ``bulk-boundary'' correlator and compare with AdS/CFT. Promisingly, they are numerically compatible given our accuracy, although further works will be needed to explore the precise connection to the AdS/CFT correspondence.
What problem does this paper attempt to address?